Stabilizing greenhouse gas concentrations at low levels: an assessment of reduction strategies and costs

Transcription

1 Climatic Change (2007) 81: DOI /s Stabilizing greenhouse gas concentrations at low levels: an assessment of reduction strategies and costs Detlef P. van Vuuren & Michel G. J. den Elzen & Paul L. Lucas & Bas Eickhout & Bart J. Strengers & Bas van Ruijven & Steven Wonink & Roy van Houdt Received: 12 December 2005 / Accepted: 9 May 2006 / Published online: 13 February 2007 # Springer Science + Business Media B.V Abstract On the basis of the IPCC B2, A1b and B1 baseline scenarios, mitigation scenarios were developed that stabilize greenhouse gas concentrations at 650, 550 and 450 and subject to specific assumptions 400 ppm CO 2 -eq. The analysis takes into account a large number of reduction options, such as reductions of non-co 2 gases, carbon plantations and measures in the energy system. The study shows stabilization as low as 450 ppm CO 2 -eq. to be technically feasible, even given relatively high baseline scenarios. To achieve these lower concentration levels, global emissions need to peak within the first two decades. The net present value of abatement costs for the B2 baseline scenario (a medium scenario) increases from 0.2% of cumulative GDP to 1.1% as the shift is made from 650 to 450 ppm. On the other hand, the probability of meeting a two-degree target increases from 0% 10% to 20% 70%. The mitigation scenarios lead to lower emissions of regional air pollutants but also to increased land use. The uncertainty in the cost estimates is at least in the order of 50%, with the most important uncertainties including land-use emissions, the potential for bio-energy and the contribution of energy efficiency. Furthermore, creating the right socio-economic and political conditions for mitigation is more important than any of the technical constraints. 1 Introduction Climate change appears to be among the most prominent sustainability problems of this century. IPCC s Third Assessment Report concludes that earth s climate system has demonstrably changed since the pre-industrial era and that without climate policy responses changes in the global climate are likely to become much larger, with expected increases in global temperature in the period ranging from 1.4 to 5.8 C (IPCC 2001). Article 2 of the United Nations Framework Convention on Climate Change (UNFCCC) states D. P. van Vuuren (*) : M. G. den Elzen : P. L. Lucas : B. Eickhout: B. J. Strengers : B. van Ruijven : S. Wonink : R. van Houdt Netherlands Environmental Assessment Agency (MNP), P.O. Box 303, 3720 AH Bilthoven, The Netherlands Detlef.van.Vuuren@mnp.nl

2 120 Climatic Change (2007) 81: as its ultimate objective: Stabilization of greenhouse gas (GHG) concentrations in the atmosphere at a level that would prevent dangerous anthropogenic interference with the climate system. However, what constitutes a non-dangerous level is an open question, as this depends on all kinds of uncertainties in the cause-effect chain of climate change and on political decisions about the risks to be avoided. Some of the recent literature suggests that climate risks could already be substantial for an increase of 1 3 C compared to pre-industrial levels (see Corfee et al 2005; ECF and PIK 2004; Leemans and Eickhout 2004; Mastandrea and Schneider 2004; MNP 2005; O`Neill and Oppenheimer 2002). As one of the political actors, the EU has adopted the climate policy goal of limiting the temperature increase to a maximum of 2 C compared to pre-industrial levels (EU 1996, 2005). However, uncertainties still allow for other interpretations of what constitutes dangerous climate change in the context Article 2. Actors may, in their interpretation, weigh factors like the risks of climate change adaptation costs and limits and the costs and effectiveness of mitigation action. Apart from the temperature target, the required level of emission reduction also depends on the uncertain relationship between atmospheric GHG concentrations and temperature increase, in other words climate sensitivity. Several probability-distribution functions (PDF) for climate sensitivity have been published in recent years, each indicating a broad range of values for climate sensitivity that still have a reasonable likelihood (for example Murphy 2004; Wigley and Raper 2001). Several authors indicated that these PDFs can be translated into a risk approach towards climate change (Azar and Rodhe 1997; den Elzen and Meinshausen 2005; Hare and Meinshausen 2004; Meinshausen 2006; Richels et al. 2004; Yohe et al. 2004). These studies show that a high degree of certainty in terms of achieving a 2 C temperature target is likely to require stabilization at low GHG concentration (for instance a probability greater than 50% requires stabilization at least below 450 ppm CO 2 - eq. 1 ). The stabilization of GHG concentrations at such a low level will require drastic emission reductions compared to the likely course of emissions in the absence of climate policies. Even for more modest concentration targets such as 650 ppm CO 2 -eq., emissions in 2100 will generally need to be reduced by about 50% compared to probable levels in the absence of a climate policy (IPCC 2001). A large number of scenario studies have been published that aim to identify mitigation strategies for achieving different levels of GHG emission reductions (see among others Hourcade and Shukla 2001; Morita and Robinson 2001). However, most of these studies have focused on reducing only the energy-related CO 2 emissions, and disregarded abatement options that reduce non-co 2 gases and the use of carbon plantations. Furthermore, the number of studies looking at stabilization levels below 550 ppm CO 2 -eq. is very limited. A few studies exist that explore the feasibility to stabilize CO 2 alone at ppm CO 2 ; the lowest multi-gas stabilization studies in literature focus on 550 ppm CO 2 -eq. (see Section 2). This implies that very little information exists on mitigation strategies that could stabilize GHG concentrations at the low levels required to achieve a 2 3 C temperature target with a high degree of certainty. As a matter of fact, even the number of studies looking at stabilizing at 550 ppm CO 2 -eq. is far lower than for higher stabilization targets (see Morita et al. 2000; Swart et al. 2002). Finally, most earlier studies have not 1 CO 2 equivalence expresses the radiative forcing of other anthropogenic radiative forcing agents in terms of the equivalent CO 2 concentration that would result in the same level of forcing. In this paper, the definition of CO 2 -eq. concentrations includes the Kyoto gases, tropospheric ozone and sulphur aerosols.

3 Climatic Change (2007) 81: considered the more recent mitigation options being discussed in the context of ambitious emission reduction, such as hydrogen and carbon capture and storage (CCS) (Edmonds et al. 2004; IEA 2004a; IPCC 2005). Given current insights into climate risks and the state of the mitigation literature, there is a very clear and explicit need for comprehensive scenarios that explore different long-term strategies to stabilize GHG emissions at low levels (Metz and van Vuuren 2006; Morita and Robinson 2001). This paper explores different multi-gas stabilization scenarios for concentration levels for which no scenarios are currently available (below 550 ppm CO 2 -eq.). In order to study the impact of different stabilization levels, we have chosen to explore scenarios for a range ofconcentrationslevels(i.e.,650,550and450ppmco 2 -eq.) and under specific assumptions (400 ppm CO 2 -eq.). As such, the study also goes beyond our own research that did not cover stabilization scenarios below 550 ppm CO 2 -eq. (van Vuuren et al. 2006b). 2 The paper adds to the existing literature in an important way by exploring pathways to those GHG stabilization levels required for achieving global mean temperature change targets of 2 3 C with a high degree of certainty. We specifically focus on the following questions: & & & What portfolios of measures could constitute promising strategies for stabilizing GHG concentrations at 650, 550 and 450 ppm CO 2 -eq. and below? What are the cost levels involved in such strategies and what are the implications for the energy sector, investment strategies and fuel trade? How do uncertainties in the potentials and costs of various options play a role in terms of the costs and selection of a portfolio of measures? The focus of this paper will be on mitigation strategies, abatement costs and climate consequences from a global perspective. In a related paper, we focus on the regional costs and abatement strategies 3 (den Elzen et al. 2006). For costs, we consider direct abatement costs due to climate policy and do not capture macro-economic costs; for benefits we focus on the impact on global mean temperature and co-benefits for air pollutants. In our analysis, we deliberately use an integrated approach, dealing with a wide range of issues that are relevant in the context of stabilization scenarios including land use consequences and changes in the energy system. Although several of these issues have been studied earlier for single stabilization scenarios, here we would like to see how they are related to the GHG stabilization level. The analysis was conducted using the IMAGE 2.3 model framework, including the energy model TIMER 2.0 coupled to the climate policy model FAIR-SiMCaP (for model description, (see Section 3). A similar framework (using FAIR instead of FAIR-SiMCaP) has been used earlier to study mitigation strategies, for example in the context of EU climate policy targets (Criqui et al. 2003; van Vuuren et al. 2003). This model framework was designed to provide a broad description of the issues involved in the chain of events causing climate change. It covers a broad range of emission sources (and therefore abatement options), covering not only the energy sector but also land use, forestry, and industry. It is therefore suitable for studying the type of mitigation strategies required to stabilize radiative forcing from GHG and for studying the possible environmental and economic consequences of such strategies. We use this framework to explore stabilization 2 Earlier we published emission profiles that would lead to stabilization at low GHG concentration levels, but that study did not look into the question how these emission profiles could be reached (den Elzen and Meinshausen 2005). 3 Regional costs also depend on possible agreements about regional reduction targets and they therefore constitute a separate topic that cannot be dealt with in the context of this paper.

4 122 Climatic Change (2007) 81: strategies based on three different baseline scenarios, i.e., updated implementations of the IPCC SRES B2, B1 and A1b scenarios. We perform an extensive sensitivity analysis for the different options to map out some of the main uncertainties. The paper is structured as follows. First, we provide a brief overview of earlier work on stabilization scenarios. We then explain the methods used to develop the new scenarios before discussing the first results from our three default scenarios and the associated benefits and cobenefits. Next, we present the results of our uncertainty analysis and also address the question of whether it is possible to reduce emissions to levels even lower than 450 ppm CO 2 -eq. After that, we compare our results to earlier studies and examine the implications of the uncertainties that have been identified. Finally, we present our overall findings. 2 Earlier work on stabilization scenarios A large number of scenario studies have been published that have explored global mitigation strategies for stabilizing GHG concentrations. A recent inventory estimated the number of published GHG emission scenarios at a few hundred, although a large majority of these are baseline scenarios (scenarios that do not take the effect of climate policy into account) (NIES 2005). 4 In the literature on mitigation scenarios, there are a number of recurring themes. These include: & & & & The issue of stabilization targets and overshoot; The identification of overall cost levels of stabilization; The issue of timing (early action or delayed response), partly in relation to technology development; The role of individual technologies and mitigation measures. In this paper, we will briefly discuss the available literature and indicate how these issues are handled. The IPCC Third Assessment Report (TAR) (Hourcade and Shukla 2001; Morita and Robinson 2001) provides an overview of the stabilization scenarios as available at that time. On the issue of stabilization targets, many studies in the past have focused on stabilizing CO 2 concentration levels. Consistent with this, new multi-gas studies mostly focus on the comparable measure of stabilizing radiative forcing (expressed in W/m 2 or CO 2 -eq.) (van Vuuren et al. 2006c). Alternatively, some studies look at temperature increase targets (as they are more directly related to impacts). One implication of using a temperature target, however, is the higher level of uncertainty relating to mitigation action (Matthews and van Ypersele 2003; Richels et al. 2004). Another issue is that staying below a certain temperature level with a specific likelihood can either be achieved by (a) stabilizing at a certain radiative forcing level or by (b) peaking at somewhat higher levels, immediately followed by a reduction of the forcing level ( overshoot scenarios ). The second 4 It is possible to distinguish between scenarios and emission pathways. Emission pathways focus solely on emissions, whereas scenarios represent a more complete description of possible future states of the world. The literature distinguishes between baseline, and mitigation or stabilization scenarios. The first category includes scenarios without explicit new climate policies. These scenarios do, however, need to assume policies in other fields than climate policy, and these may still unintentionally have a significant impact on GHG emissions (e.g., other environmental policies, trade policies). Mitigation scenarios (or climate policy scenarios) purposely assume climate policies to explore their impact. Stabilization scenarios are a group of scenarios that include mitigation measures intended to stabilize atmospheric GHG concentrations.

5 Climatic Change (2007) 81: strategy prevents some of the temperature increase that will occur in the longer term (den Elzen and Meinshausen 2005; Meinshausen 2006; Wigley 2003). In general, these overshoot scenarios will involve lower costs than the corresponding stabilization scenarios. For the lower stabilization levels, overshoot scenarios are the only feasible scenarios since current concentrations have either already passed these levels, or will do so in the very near future. In broad terms, the current scenario literature covers stabilization levels from 750 to 450 ppm CO 2 for CO 2 -only studies. There are only a few studies that have looked into stabilizing concentrations at lower concentration levels. Exceptions include the work of Nakicenovic and Riahi (2003), Azar et al. (2006), and Hijoka et al. (2006). These studies show that, in principle, low stabilization levels (below 450 ppm CO 2 ) can be achieved at mitigation costs in the order of 1% 2% of GDP. However, these studies started from relatively low-emission baseline scenarios. In multi-gas studies, the range is actually much more limited, with studies typically only looking at 650 ppm CO 2 -eq. (van Vuuren et al. 2006c; Weyant et al. 2006). The lowest scenarios currently found in the literature aim at 550 ppm CO 2 -eq. (Criqui et al. 2003; van Vuuren et al. 2006b) and these only give a very low level of probability to limit temperature increase to less than 2 C. For a range of probability-distribution functions (PDF), Hare and Meinshausen (2004) estimated the probability to be about 0% 30%. The probability of staying within 2.5 C is 10% 50%. A 50% probability (on average) of staying within 2 C is obtained for 450 ppm CO 2 -eq. The only multi-gas studies in the literature that are currently exploring the consequences of aiming for such low stabilization levels are emission pathway studies that do not specify the type of mitigation measures leading to the required emissions reductions (den Elzen and Meinshausen 2005; Meinshausen 2006; Meinshausen et al. 2006). Different metrics are used for the costs of climate policies. Partial equilibrium models (such as energy system models) generally report costs as increased energy system costs or abatement costs (these are annual costs that can be expressed as percentages of GDP). General equilibrium models, by contrast, generally report reductions of GDP or private consumption relative to the baseline scenario. For the stabilization scenarios analyzed in TAR, the assessment found very small costs for stabilizing at 750 ppm but stated typical GDP losses of 1% 4% for 450 ppm (Hourcade and Shukla 2001). Costs were found to be a function of the GHG stabilization level and the baseline emission scenario. This implies that socio-economic conditions, including policies outside the field of climate policy, are just as important for stabilization costs as climate policies. The issue of the timing of the abatement effort was initiated by Hamitt et al. (1992) and later by Wigley et al. (1996). Wigley et al. (1996) argued that their scenarios, which postponed abatement action compared to earlier pathways developed by IPCC, were more cost-effective because of the benefits of technology development, more CO 2 absorption by the biosphere and oceans, and discounting future costs. Their arguments were confirmed in the analysis of the EMF-14 (energy modeling forum) study (as reported by Hourcade and Shukla 2001). Other authors, however, responded that this conclusion depends on the assumptions about discounting, technological change, inertia and uncertainty (Azar 1998; Azar and Dowlatabadi 1999; Ha-Duong et al. 1997). For low-range concentration targets, den Elzen and Meinshausen (2005) reported that delaying the peak in global emissions beyond 2020 leads to very high reduction rates later in the century and therefore to probable high costs. Assuming induced technology change (instead of exogenous technological progress simply as function of time) and explicit capital turnover rates could lead to a preference for early action, or at least a spread of the reduction effort over the century as a whole (see also van Vuuren et al. 2004). The debate about optimal timing is still ongoing. Yohe et al. (2004) recently showed that hedging strategies (i.e., cost-optimal reduction

6 124 Climatic Change (2007) 81: pathways incorporating the risk of more, or less, stringent action later in the century if new evidence comes in) to deal with uncertainties may lead to relatively early reduction pathways leaving as many options open as possible (Berk et al. 2002). Recently, a lot of attention has been paid to extending the number of reduction options considered in scenario analysis. One possibility is the inclusion of non-co 2 GHGs. The energy modeling forum (EMF-21) performed a model comparison study, showing that extending the reduction options from CO 2 only to include other GHGs can reduce costs by about a third (van Vuuren et al 2006c; Weyant et al. 2006). Recent publications also put forward several new technologies that could be pivotal in mitigation strategies. First of all, CCS could play an important role in reducing GHG emissions in the power sector. This technology could become cost-effective at emission permit prices of around US$/ tc (IPCC 2005) and therefore reduce mitigation costs considerably (Edmonds et al. 2004; IEA 2004a). Recent work on hydrogen as an energy carrier has shown that hydrogen may also reduce mitigation costs but this conclusion depends very much on the assumption of technology development (Edmonds et al. 2004). Bio-energy in combination with CCS could be an attractive technology if very ambitious stabilization targets are adopted (Azar et al. 2006). Finally, the debate is still ongoing about whether accounting for technology change (induced learning vs exogenous assumptions) in itself results in different conclusions about optimal climate policies. Some studies claim that induced technological change leads to very significant cost reductions and justifies a preference for early action (Azar and Dowlatabadi 1999; Barker et al. 2005). Others report fewer benefits and/or no impact on timing (Manne and Richels 2004). What are the implications of the current state of knowledge for this study? The most important aim of this study is to determine whether a multi-gas approach can be used to achieve the stabilization of GHG concentration at lower levels than those usually considered in mitigation studies. Our scenarios, based on the emission pathways developed by den Elzen and Meinshausen (2005) and den Elzen et al. (2007) should be characterized as mediumterm pathways (since they are neither early nor delayed). In terms of the objective of climate policy, we focus on the stabilization of concentration (and thus not temperature) to increase the comparability with other studies. den Elzen et al. (2007) indicate how the results of the emission pathways compare to alternative peaking scenarios. In view of the debate about new mitigation options, the model framework used in this study covers a large range of mitigation options (such as non-co 2, CCS, carbon plantations, hydrogen, bioenergy, nuclear, solar and wind power), and several technologies are described in terms of induced technological change. The aim of this study is to identify a portfolio of measures that contribute to the reduction of emissions with the aim of achieving the selected concentration targets, and to assess the costs associated with this portfolio. Given the major uncertainties involved in each of the mitigation options, we will analyze how some of these uncertainties impact the overall results. 3 Methodology 3.1 Overall methodology For the construction of the stabilization scenarios, we use an interlinked modelling framework consisting of the IMAGE 2.3 integrated assessment model, which includes the TIMER 2.0 energy model (de Vries et al. 2001) coupled to the climate policy model FAIR-SiMCaP

7 Climatic Change (2007) 81: TIMER 2.0 Baseline 1 Biofuel potential Emissions and biofuel demand IMAGE 2.3 Baseline 1 Baseline emissions and energy MAC curves Baseline emissions and CP MAC curves 2 FAIR-SimCap Emission profile 2 FAIR 2.0 Costs / reductions Non-CO2 MAC curves 2 Emission reductions and marginal price Emission reductions and CP demand 3 TIMER 2.0 Mitigation scenario Emissions and biofuel demand 3 IMAGE 2.3 Mitigation scenario Fuel trade and Land-use and energy impacts climate impacts Fig. 1 Linkage and information flows of the applied modelling framework (note CP Carbon plantations). Numbers in figure are explained in the text (den Elzen and Lucas 2005; den Elzen and Meinshausen 2005). 5 These models have been linked for the purposes of this analysis in a way similar to that described earlier by van Vuuren et al. (2003), 6 as shown in Fig. 1. The Appendix provides additional information on the different models used. The IMAGE 2.3 model is an integrated assessment model consisting of a set of linked and integrated models that together describe important elements of the long-term dynamics of global environmental change, such as air pollution, climate change, and land-use change. IMAGE 2.3 uses a simple climate model and a pattern-scaling method to project climate change at the grid level. At the grid level, agriculture is described by a rule-based system driven by regional production levels. Finally, natural ecosystems are described by an adapted version of the BIOME model. The global energy model, TIMER 2.0, a part of the IMAGE model, describes primary and secondary demand for, and production of, energy and the related emissions of GHG and regional air pollutants. The FAIR-SiMCaP 1.1 model is a combination of the multi-gas abatement-cost model of FAIR 2.1 and the pathfinder module of the SiMCaP 1.0 model. The FAIR cost model distributes the difference between baseline and global emission pathways using a least-cost approach involving regional marginal abatement cost (MAC) curves 5 IMAGE 2.3 is an updated version of IMAGE 2.2, the difference being the possibility of exploring impacts of bio-energy and carbon plantations. TIMER 2.0 is an updated version of TIMER 1.0. The main differences are additions with respect to hydrogen, bio-energy and modelling of the electric power sector. The FAIR- SIMCAP model is the combination of the climate policy support tool FAIR and the SiMCaP model. 6 In the present framework, FAIR-SIMCAP is used for the calculations of the global emission pathways instead of the IMAGE 2.2 model.

8 126 Climatic Change (2007) 81: for the different emission sources (den Elzen and Lucas 2005). 7 The SiMCaP pathfinder module uses an iterative procedure to find multi-gas emission paths that correspond to a predefined climate target (den Elzen and Meinshausen 2005). Calculations in all three main models are done for 17 regions 8 of the world. The overall analysis consists of three major steps (Fig. 1): 1. Both the IMAGE and the TIMER model are used to construct a baseline emission scenario. Furthermore, the TIMER model yields the potentials and abatement costs of reducing emissions from energy-related sources, while the IMAGE model provides the potentials and abatement costs associated with carbon plantations. 2. The FAIR-SiMCaP 1.1 model is used to develop global emission pathways that lead to a stabilization of the atmospheric GHG concentration. The concentration calculations are done using the MAGICC 4.1 model that is included in the FAIR-SiMCaP 1.1 model. (Wigley and Raper 2001). The FAIR model distributes the global emission reduction from the baseline across the different regions, gases and sources in a cost-optimal way, using the marginal abatement costs curves. It is assumed that the gases are substituted on the basis of GWPs, an approach consistent with climate policies under the Kyoto Protocol and the USA domestic climate policy (White-House 2002). Furthermore, the model calculates the international permit price, the regional emission reductions, and the global and regional costs of emission reductions. 3. The IMAGE/TIMER model implements the changes in emission levels resulting from the abatement action (emission reductions) and the permit price, as determined in the previous step, to develop the final mitigation scenario (emissions, land use, energy system). Furthermore, the environmental impacts are assessed using the climate model of IMAGE. In our analysis, we assume that reductions can be distributed across all 17 regions costoptimally from 2013 onwards. This implies the presence of some form of international mechanism that justifies this least-cost assumption, such as emission trading. 3.2 Baseline emissions The baseline scenarios used in this study are based on IPCC-SRES scenarios (Nakicenovic et al. 2000b). This set of baseline scenarios explores different possible pathways for GHG emissions and can roughly be categorized along two dimensions: the degree of globalization vs regionalization, and the degree of orientation towards economic objectives as opposed to an orientation towards social and environmental objectives. In 2001, the IMAGE team published detailed elaborations of these scenarios (IMAGE-team 2001a, b). Although the scenarios are still broadly consistent with the literature, new insights have emerged for some parameters. For instance, population scenarios and economic growth assumptions for low-income regions are now generally lower than assumed in SRES (van Vuuren and O_Neill 2006). Against this background, a set of updated IMAGE scenarios was developed (see Fig. 2). Here, we use the B2 scenario as the main baseline scenario, with the A1b and B1 scenarios being used to show the impacts of different baseline assumptions. 7 Marginal Abatement Cost (MAC) curves reflect the additional costs of reducing the last unit of CO 2 -eq. emissions. 8 Canada, USA, OECD-Europe, eastern Europe, the former Soviet Union, Oceania and Japan; Central America, South America, northern Africa, western Africa, eastern Africa, southern Africa, Middle East and Turkey, south Asia (incl. India), South-East Asia and east Asia (incl. China) (IMAGE-team 2001a).

9 Climatic Change (2007) 81: Population 20 GDP Population (billion) Index (2000 = 1) Marker A1 A2 B1 B IMAGE Final energy consumption 40 CO2 A1 A B1 B2 Use (EJ) 1000 CO2 (GtC) Fig. 2 Driving forces and fossil fuel CO 2 emissions of the IMAGE 2.3 SRES scenarios in comparison to the IPCC SRES marker scenarios (Nakicenovic et al. 2000a) The new implementation of B2 focuses explicitly on exploring the possible trajectory of greenhouse gas emissions on the basis of medium assumptions for the most important drivers (population, economy, technology development and lifestyle). In terms of its quantification, the B2 scenario roughly follows the reference scenario of the World Energy Outlook 2004 for the first 30 years (IEA 2004b). After 2030, economic growth converges to the B2 trajectory of the previous IMAGE scenarios (IMAGE-team 2001a). The long-term UN medium population projection is used for population (UN 2004). The A1b scenario, by contrast, represents a world with fast economic growth driven by further globalization and rapid technology development. As the scenario that materialintensive lifestyles are adopted globally, energy consumption grows rapidly. The B1 scenario describes a world characterized by strong globalization in combination with environmental protection and a reduction of global inequality. It assumes the use of very efficient technologies, resulting in relatively low energy use. The assumptions for population and economic growth in the A1 and B1 scenarios have been taken from, respectively, the Global Orchestration and Technogarden scenarios of these (MA 2006). In all three scenarios, trends in agricultural production (production levels and yields) are also based on the millennium ecosystem scenarios, which were elaborated for these parameters by the IMPACT model (Rosegrant et al. 2002). All other assumptions are based on the earlier implementation of the SRES scenarios. 3.3 Assumptions in the different subsystems and marginal abatement costs We adopt a hybrid approach to determine the abatement efforts among the different categories of abatement options. First, the potential for abatement in different parts of the system

10 128 Climatic Change (2007) 81: (energy, carbon plantations, and non-co 2 ) is translated into aggregated baseline- and timedependent MAC curves. These curves are than used in the FAIR model to distribute the mitigation effort among these different categories and to determine the international permit price. The corresponding reduction measures at the more detailed level are determined by implementing the permit price in the different expert models for energy (TIMER) and carbon plantations (IMAGE). For instance, in the case of energy, the TIMER model results in a consistent description of the energy system under the global emission constraint set by FAIR-SiMCaP. The TIMER, IMAGE and FAIR-SiMCaP models have been linked so that output of one model is the input of the second model (see Fig. 1). In addition, also the model-specific assumptions in the different models have been harmonized. In most cases, this was done on the basis of the storyline of the different scenarios being implemented. For example, technology development is set low for all parameters in the different models in the A2 scenario. The same holds for other driving forces. In terms of land use, both carbon plantations and bio-energy calculations start with the same land-use scenario (implementation factors prevent them using the same land) and the same land price equations. A 5% per year social discount rate is used to calculate the net present value for the mitigation scenarios. In the energy system, investment decisions are compared using a 10% per year discount rate, which provides a better reflection of the medium term investment criteria used in making such investments. Table 1 summarizes some of the assumptions made. All costs are expressed in 1995 US$ Energy The TIMER MAC curves (used by the FAIR model) are constructed by imposing an emission permit price (carbon tax) and recording the induced reduction of CO 2 emissions. 9 There are several responses in TIMER to adding an emission permit price. In energy supply, options with high carbon emissions (such as coal and oil) become more expensive compared to options with low or zero emissions (such as natural gas, CCS, bio-energy, nuclear power, solar and wind power). The latter therefore gain market share. In energy demand, investments in efficiency become more attractive. The induced reduction of CO 2 emissions is recorded for sight-years from 2010 to 2100 (in 10-year steps). Two different permit price profiles were used to explore responses: one that assumes a linear increase from 2010 to the permit price value in the sight year ( linear price MAC ) and one that reaches the maximum value 30 years earlier ( block price MAC ). The second profile results in more CO 2 reductions because the energy system has more time to respond. Depending on the pathway of the actual permit price in the stabilization scenario, FAIR combines the linear price MAC curves and the block price MAC curves. 10 In this way, it is possible to take into account (as a first-order approximation) the time pathway of earlier abatement. In the baseline, stricter investment criteria are used for investments in energy efficiency than for investments in energy supply. Investments in energy efficiency are made only if the apparent average pay-back-time is less than three years (for industry) or two years (other 9 The tax is intended to induce a cost-effective set of measures and is in the model equivalent to a emission permit price. In the rest of the paper, we will use the term (emission) permit price. It should be noted that in reality, the same set of measures as induced by the permit price can also be implement by other type of policies. 10 FAIR looks 30 years back in time and, by comparing the tax profile in that period to the one assumed in the tax profiles used in TIMER, constructs a linear combination of the two types of response curves. A rapidly increasing tax in FAIR will lead to the use of the linear tax, while a more constant tax level in FAIR will imply the use of the block tax.

11 Climatic Change (2007) 81: Table 1 Default assumptions for various reduction options and the alternative assumptions used in the sensitivity analysis Mitigation option Pessimistic assumption Base case Optimistic assumption Carbon plantations Non-CO 2 Carbon uptake reduced by 25% + implementation factor reduced to 30% 20% increase of costs; 20% decrease of potential Implementation factor 40% (i.e., 40% of maximum potential is used) Expert judgment as described in Lucas et al. (2005). Total reduction potential of non-co 2 gases slightly above 50% Hydrogen No hydrogen penetration Default assumptions lead to hydrogen penetration by the end of the century Efficiency improvement Bio-energy Climate policies do not lead to removal of implementation barriers for efficiency Less available land for bio-energy (50% less) Climate policies lead to some removal of implementation barriers for efficiency Carbon uptake increased by 25% + implementation factor increased to 50% 20% decrease of costs; 20% increase of potential Optimistic assumptions for fuels cells and H 2 production costs (10% reduction of investment costs) lead to penetration around 2050 Climate policies lead to full removal of implementation barriers for efficiency Bio-energy can also be used in combination with CCS technology Technology development No climate policy-induced learning Climate policy-induced learning Carbon capture and storage No carbon capture and storage Medium estimates for CCS storage potential (see Table 5) Nuclear Nuclear not available as mitigation option Nuclear available as mitigation option Emission Emission trading Full emission trading trading restricted due to transaction costs of 15$/tC Land use Agricultural yields do not improve as fast (following MA s Order from strength scenario) Medium yield increases (following MA s adaptive mosaic scenario) Agricultural yields do not improve as fast (following MA s global orchestration scenario) Baseline IMAGE 2.3 A1b IMAGE 2.3 B2 IMAGE 2.3 B1 All All above, excluding land use and baseline All above, excluding land use and baseline All above, excluding land use and baseline Not for all options both more pessimistic and more optimistic assumptions were tested.

12 130 Climatic Change (2007) 81: sectors) (see de Beer 1998). 11 In low-income countries, we assume that lower efficiency levels are caused by less strict apparent pay-back-time criteria (de Vries et al. 2001). The criteria used in energy supply (based on a 10% discount rate and the economic life time depending on the type of technology applied) corresponds more-or-less to a pay-back time of six to seven years. The difference between demand and supply investment criteria is based on historical evidence (barriers to demand-side investments include lack of information, more diffuse investors, higher risks and lack of capital). Under climate policies, investments into energy efficiency could therefore form a very cost-effective measure if these barriers can be overcome. In our calculations, we assume that this is partly the case as a result of (1) an increase in attention for ways to reduce carbon emissions (leading to more information) and (2) the availability of capital flows, including flows to developing countries, that could possibly result from carbon trading (or other flexible mechanisms). Based on this, we assume a convergence of the pay-backtime criterion to six years as a function of the existing emission permit price with full convergence at the highest price considered, i.e., 1,000 US$/tCeq Carbon plantations The MAC curves for carbon plantations have been derived using the IMAGE model (for methodology, see (Graveland et al. 2002; Strengers et al. 2007). In IMAGE, the potential carbon uptake of plantation tree species is estimated for land that is abandoned by agriculture (using a grid), and compared to carbon uptake by natural vegetation. Only those grid cells are considered in which sequestration by plantations is greater than sequestration by natural vegetation. In the calculations, we assumed that carbon plantations are harvested at regular time intervals, and that the wood is used to meet existing wood demand. Regional carbon sequestration supply curves are constructed on the basis of grid cells that are potentially attractive for carbon plantations. These are converted into MAC curves by adding two kinds of costs: land costs and establishment costs. We found that, under the SRES scenarios, the cumulative abandoned agricultural area ranges from 725 and 940 Mha in 2100, potentially sequestering 116 to 146 GtC over the century (the term agricultural land in this paper covers both crop and pasture land). The costs of the reductions vary over a wide range Non-CO 2 gases For non-co 2, the starting point of our analysis consists of the MAC curves provided by EMF-21 (Weyant et al. 2006). This set is based on detailed abatement options, and includes curves for CH 4 and N 2 O emissions from energy- and industry-related emissions and from agricultural sources, as well as abatement options for the halocarbons. This set includes MAC curves over a limited cost range of 0 to 200 US$/tC-eq., and does not include technological improvements over time. Lucas et al. (in press) have extended this set on the basis of a literature survey and expert judgement about long-term abatement potential and costs (see also van Vuuren et al. 2006b). The long-term potential is significantly higher than current potential as a result of technology development and the removal of 11 A pay-back-time is a simple investment criterion that indicates the time-period required to earn back the original investment. Research indicates that many actors are not aware of the energy efficiency improvement measures that are available to them that have shorter pay-back-time periods than their official criterion. As a result, the average apparent pay-back-time of a sector is considerably lower than the investment criteria that are stated to be used by these actors (de Beer 1998).

13 Climatic Change (2007) 81: implementation barriers. The overall potential amounts to about 3 GtC-eq. annually (with the lion s share available below 200 US$/tC-eq.). 3.4 Emission pathways This study uses a set of global multi-gas emission pathways that meet GHG concentration stabilization targets at 450, 550 and 650 ppm CO 2 -eq. (den Elzen and Meinshausen 2005). These pathways are assumed to be technically feasible, as we calculated them using the MAC curves discussed above. In general terms, three main criteria were used when developing the pathways. First, a maximum reduction rate was assumed reflecting the technical (and political) inertia that limits emission reductions. Fast reduction rates would require the early replacement of existing fossil-fuel-based capital stock, and this may involve high costs. Secondly, the reduction rates compared to baseline were spread out over time as far as possible but avoiding rapid early reduction rates and, thirdly, the reduction rates were only allowed to change slowly over time. The selected values are based on the reduction rates of the post-sres mitigation scenarios (e.g., Swart et al. 2002) and the lower range of published mitigation scenarios (Azar et al. 2006; Nakicenovic and Riahi 2003). In the case of the 650 and 550 ppm CO 2 -eq. pathways, the resulting pathway leads to stabilization below the target level and without overshoot between 2100 and For the 450 ppm CO 2 -eq. concentration target, however, a certain overshoot (or peaking) is assumed: concentrations may first increase to 510 ppm before stabilizing at 450 ppm CO 2 -eq. before This overshoot is justified by reference to present concentration levels, which are already substantial, and the attempt to avoid drastic sudden reductions in the emission pathways presented. 4 Stabilizing GHG concentration at 650, 550, 450 ppm: central scenarios 4.1 Emission pathways and reductions Under the central baseline, B2, worldwide primary energy use nearly doubles between 2000 and 2050 and increases by another 35% between 2050 and Most of this growth occurs in non-annex I regions (about 80%). Oil continues to be the most important energy carrier in the first half of the century, with demand being mainly driven by the transport sector. Natural gas dominates new capacity in electric power in the first decades, but starts to be replaced by coal from 2030 onwards due to increasing gas prices. As a result, coal becomes the dominant energy carrier in the second half of the twenty-first century. Energy-sector CO 2 emissions continue to rise for most of the century, peaking at 18 GtC in Total GHG emissions 12 also increase, i.e., from about 10 GtC-eq. today to 23 GtC-eq. in 2100 (Fig. 3). Figure 3 also shows that compared to existing scenario literature; this baseline is a medium-high emission baseline. As a result of decreasing deforestation rates, CO 2 emissions from land use decrease. At the same time, CH 4 emissions, mostly from agriculture, increase. The GHG concentration reaches a level of 925 ppm CO 2 -eq., leading to an increase in the global mean temperature of 3 C in 2100 (for a climate sensitivity of 2.5 C). Figure 3a shows that, in order to reach the selected emission pathway that leads to stabilization of GHG radiative forcing at 650, 550 and 450 ppm CO 2 -eq., GHG emissions need to be reduced in 2100 by respectively 65%, 80% and 90% compared to the B2 baseline. 12 The term total GHG emissions in this report refers to all GHG covered by the Kyoto protocol: i.e., CO 2, CH 4,N 2 O, HFCs, PFCs and SF 6.

14 132 Climatic Change (2007) 81: a) B2 Baseline and mitigation profiles 40 b) B2 Baseline and other baselines Emissions (GtC-eq) B2 Baseline CO 2 Non-CO CO 2 -eq 550 CO 2 -eq Emissions (GtC-eq) IMAGE 2.3 B2 IMAGE 2.3 A1b IMAGE 2.3 B1 Grey area indicates EMF21 range CO 2 -eq Fig. 3 Global CO 2 -eq. emissions (all sources 2 ) for the B2 baseline emission and pathways to stabilization at a concentration of 650, 550 and 450 ppm CO 2 -eq. (a, left) and the B2 baseline emissions compared to alternative baselines (b, right), sources: for the EMF-21 scenarios (van Vuuren et al. 2006c; Waterloo et al. 2001) The short-term differences are even more significant: In the case of the 650 ppm CO 2 -eq. pathway, emissions can still increase slightly and stabilize at a level that is 40% above current emissions in the next three to four decades, followed by a slow decrease. In the case of the 550 ppm CO 2 -eq. pathway, however, global emissions need to peak around 2020, directly followed by steep reductions in order to avoid overshooting the 550 ppm CO 2 -eq. concentration level. For stabilization at 450 ppm CO 2 -eq., short-term reductions become even more stringent, with global emissions peaking around 2020 at a level of 20% above 2000 levels. 4.2 Abatement action in the stabilization scenarios Abatement across different gases Figure 4 shows the (cost-optimal) reduction in the mitigation scenarios in terms of different gases (upper panel). Table 2, in addition, indicates the emission levels. In the short term, in all stabilization scenarios, a substantial share of the reduction is achieved by reducing non-co 2 gases while only 10% of the reductions come from reducing energy-related CO 2 emissions (see also Lucas et al. 2005). The disproportionate contribution of non-co 2 abatement is caused mainly by relatively low-cost abatement options that have been identified for non-co 2 gases (e.g., reducing CH 4 emissions from energy production and N 2 O emissions from adipic and acidic acid industries). It should be noted that this is related to the fact that we use GWPs to determine the cost-effective mix of reductions among the different GHGs (see Section 3). Alternative approaches, e.g., long-term costs optimization under a radiative forcing target, may result to a different mix (van Vuuren et al. 2006c). After 2015, more and more reductions will need to come from CO 2 in the energy system, increasing to 85% by This shift simply reflects that non-co 2 represents about 20% of total GHG emissions and the limited reduction potential for some of the non-co 2 gases. In addition, some non- CO 2 GHGs cannot be reduced fully due to limited reduction potential (this is the case for some sources of land-use-related CH 4 but is particularly true for some of the N 2 O emission sources, see below). The proportion of non-co 2 abatement does decline somewhat further in the 450 ppm CO 2 -eq. scenario than in the 650 ppm CO 2 -eq. scenario (with the proportion being limited by the absolute non-co 2 reduction potential). More detailed analysis across the different sources shows that, for CH 4, relatively large reductions are achieved in the areas of landfills and the production of coal, oil and gas.

15 Climatic Change (2007) 81: ppm CO 2 -eq. 550 ppm CO 2 -eq. 450 ppm CO 2 -eq. a) Contribution to GHG reduction by gas Emissions (GtC-eq) Emissions (GtC-eq) Emissions (GtC-eq) b) Contribution to CO2 reduction by measure Emissions (GtC-eq) Emissions (GtC-eq) Emissions (GtC-eq) a) Contribution to GHG reduction by gas Carbon plantations F-gasses b) Contribution to CO 2 reduction by measure Fuel switch CCS N2O Biofuels CH4 Nuclear, solar & wind power CO2 Efficiency improvement Fig. 4 Emission reductions for total GHG emissions contributed by gas (upper panel; a) and for energy CO 2 emissions contribute by reduction measure category (lower panel; b) applied to stabilization scenarios at 650, 550 and 450 ppm CO 2 -eq In total, under the 450 ppm CO 2 -eq. stabilization scenario, emissions are reduced by 70% compared to the baseline. In the less stringent 650 ppm stabilization case, CH 4 emissions are halved (returning roughly to today s levels). In the case N 2 O, substantial reductions are achieved for acidic and adipic acid production (up to 70% reduction). However, in comparison to land-use related N 2 O emissions, this only represents a small source. For the Table 2 Emissions in 2000 and in 2100 for the B2 baseline and the stabilization scenarios Baseline Stabilization scenarios (ppm CO 2 -eq.) GtC-eq. CO 2 energy/industry Electricity sector Industry Buildings Transport Other Total CO 2 land use CH N 2 O F-gases Total

16 134 Climatic Change (2007) 81: Energy use (EJ) a) B2 baseline b) 450 ppm CO eq. stabilization Energy use (EJ) Nuclear Renewables Bio-energy + CCS Natural gas+ccs Oil+CCS Coal+CCS Bio-energy Natural gas Oil Coal Fig. 5 Primary energy use in the B2 baseline (left, a) and the 450 ppm CO 2 -eq. stabilization scenario (right, b). Note: Nuclear, solar, wind and hydro power have been reported at a virtual efficiency of 40%; bioenergy includes traditional biofuels; renewables include hydro, solar and wind power land-use-related N 2 O sources, emission reduction rates are smaller. As a result, total N 2 O emission reductions in the most stringent scenario amount to about 35% compared to baseline. In the most stringent case, emissions of halocarbons are reduced to almost zero for the group as a whole. In the other two scenarios, considerable reduction rates are still achieved. The use of carbon plantations contributes about 0.9 GtC annually to the overall mitigation objective in 2100 in the 450 ppm CO 2 -eq. scenario but less in the other two scenarios (0.5 and 0.25 GtC annually). In all three scenarios, East Asia, South America and the former Soviet Union together account for more than 50% of the carbon plantation mitigation effort (regional detail is not shown in figures but can be found in Strengers et al. (2007). The trees used vary according to the location and include Populus Nigra (East Asia and Europe), Picea Abies (Canada, USA and former USSR) and E. Grandis (South America, central Africa and Indonesia). In all three scenarios, high sequestration rates (more than 0.1 GtC annually) are achieved only after due to limited land availability early on. Some of the mitigation by carbon plantations can be achieved at relatively low-costs and form a substantial part of the potential used in the 650 ppm CO 2 -eq. stabilization scenario. The potential of carbon plantations does depend more on external assumptions (e.g., the implementation fraction) than on the stabilization target Abatement action in the energy system Figure 5 shows that the climate policies required to reach the stabilization pathways lead to substantial changes in the energy system compared to the baseline scenario (shown for 450 ppm CO 2 -eq.). These changes are more profound when going from 650 to 450 ppm CO 2 -eq. In the most stringent scenario, global primary energy use is reduced by around 20%. Clearly, the reductions are not similar for the different energy carriers. The largest reductions occur for coal, with the remaining coal consumption being primarily used in electric power stations using CCS. There is also a substantial reduction for oil. Reductions for natural gas are less substantial, while other energy carriers in particular solar, wind and nuclear-based electricity and modern biomass gain market share. 13 The largest reduction in the energy sector results from changes in the energy supply (Fig. 4; lower panel). Some changes stand out. First of all, under our default assumptions, 13 Modern biomass includes gaseous or liquid fuels produced from plants or trees. It differs from traditional biomass (gathered wood, straw, dung, charcoal, etc).

17 Climatic Change (2007) 81: CCS mainly in the power sector accounts for a major proportion of the emission reductions (up to a third of the reductions in energy-related CO 2 emissions). As a result, large amounts of CO 2 are stored. In the 650 ppm case, 160 GtC, or about 2 GtC annually on average, needs to be stored, mainly in empty gas and oil fields. In the 550 and 450 cases, these numbers are 250 and 300 GtC, or about 3 GtC annually. Here, we use medium estimates of storage capacity (around 1,000 GtC) but estimates in the low range are in the order of 100 GtC (Hendricks et al. 2002). In the more densely populated regions, we find that under our medium assumptions reservoirs from depleted fossil fuel resources will be filled near the end of the century so that these regions will also use aquifers as a storage option. 14 The decreasing reservoir capacity will lead to slightly higher costs. It should be noted that CCS technology still has to be proven in large scale application and aquifer capacity is uncertain. Bio-energy use also accounts for a large proportion of the emission reductions. In the baseline scenario of this study about 200 EJ of bio-energy are used. In the most stringent stabilization scenario, bio-energy use increases to 350 EJ. In terms of crops, the bio-energy is produced from a mixture of crops (sugar cane, maize and woody bio-energy depending on the region). The use of bio-energy requires land where, in the baseline, there would be regrowing natural vegetation sequestering carbon. The decrease in carbon sequestration by bio-energy production compared to natural vegetation regrowth amounts to about 1 5 kg C per GJ of bio-energy produced, depending on the region and biome (this number represents the annual average across the whole scenario period, by taking the cumulative bio-energy production and the cumulative difference in carbon uptake between the land used for bioenergy production and the original vegetation). This compares to standard emission factors of 25 kg C per GJ for coal, 20 kg C per GJ for oil and 15 kg C per GJ for natural gas. The contribution shown in Fig. 4 indicates the net contribution. Solar, wind and nuclear power also account for a considerable proportion of the required reductions. In our baseline scenario, the application of renewables (i.e., hydro, wind and solar power) is considerably larger than that of nuclear power (based on current policies and costs). In the mitigation scenario both categories increase their market share. For hydropower, we assumed no response to climate policy (given the fact that in the baseline most regions are already approaching their maximum potential levels and investments into hydropower are often related to other objectives than energy alone). As a result of their intermittent character, the contribution of solar and wind power is somewhat limited by a declining ability to contribute to a sufficiently reliable electric power system at high penetration rates. As a result, in the model the increase in nuclear power compared to the baseline is larger than that of renewables. The finding that under climate policy, nuclear power could become a competitive option to produce electric power is consistent with several other studies (MIT 2003; Sims et al. 2003). However, more flexible power systems, different assumptions on the consequences of intermittency for renewables, the development of storage systems, technological breakthroughs or taking account of public acceptance of nuclear power could easily lead to a different mix of nuclear power, solar and wind power and CCS technologies (and still lead to a similar reduction rate). Energy efficiency represents a relatively important part of the portfolio early on in the century but a much smaller share compared to baseline later on. The main reason for the 14 In our analysis we have used the reservoir estimates as estimated by Hendriks et al. (2002), including their estimates for aquifers. Hendriks et al. (2002) restricted the potentially available storage capacity in aquifers severely based on safety requirements for storage. Still, one might argue that the reservoir estimates for aquifers are more uncertain as those for (empty) fossil fuel reservoirs.

18 136 Climatic Change (2007) 81: decreasing impact is that costs reductions of zero carbon energy supply options reduces the effectiveness of energy efficiency measures. In addition, the fact that energy efficiency will be closer to the technology frontier in many parts of world will slow down further improvement. Globally, energy use is reduced in 2100 by about 10% in the 650 ppm case and about 20% in the 450 ppm case. The contribution of efficiency does vary strongly by region and over time. In western Europe, for instance, in the model the annual rate of real efficiency improvement in the baseline is about 1.1% per year in the first half of the century, and 0.8% per year over the century as a whole (these numbers refer to the underlying efficiency indicators in the model; not the energy intensity (energy over GDP) that improves somewhat faster due to structural change). The increased energy prices under climate policies in combination with the reduction of investment barriers could raise the numbers to 1.5% and 1.0% per year respectively in the 450 ppm CO 2 -eq. scenario. In India, climate policy could have a much larger impact. Here, baseline efficiency improvement is assessed at 2.2% per year in the first 40 years and 1.8% per year over the century. Climate policies could push up these numbers to 2.9% and 2.1% per year respectively. An alternative way to look at these data is to use the Kaya indicators of energy intensity (GJ/$) and the carbon factor (kg C/GJ) (Kaya 1989). Under the baseline scenario, energy intensity improves significantly by about 70% worldwide between 2000 and The carbon factor remains virtually constant (in line with historic trends). It is only in the last decades that some decarbonization occurs as high oil prices induce a transition to bio-energy. This implies that, in the baseline scenario, energy intensity improvement is the main contributor to decreasing the ratio between CO 2 emissions and GDP growth (kg C/GDP). In the mitigation scenarios, the rates increase for both energy intensity and carbon factor improvement. While the contribution of the two factors to emission reductions compared to baseline levels is about the same in 2020 (this can be seen in Fig. 6 since the mitigation scenario 2020 points are moved parallel to the diagonal in the figure compared to the baseline scenario points), changes in the carbon factor compared to baseline (in other words: changes in energy supply) in 2050 and 2100 contribute much more to lower emission levels than energy intensity. Under the 450 ppm scenario, the carbon factor decreases by about 85% compared to baseline by the end of the century. 4.3 Costs Abatement costs As costs measures, we will focus on marginal permit prices and abatement costs. The latter are calculated on the basis of the marginal permit prices and represent the direct additional costs due to climate policy, but do not capture macro-economic costs (nor the avoided damages of climate change). Figure 7 shows that the scenarios involving stabilization at 650 and 550 ppm CO 2 -eq. ppm are characterized by a rather smooth increase in the marginal price followed by a drop by the end of the century. The latter is caused by a fall in emissions in the baseline and further cost reductions in mitigation technologies (in particular, hydrogen fuel cells start entering the market by this time, allowing for reductions in the transport sector at much lower costs). For the 450 ppm stabilization scenario, the marginal price rises steeply during the first part of the century reaching a marginal price of over 600 US$/tC-eq. by 2050 and finally stabilizes at 800 US$/tC-eq. by the end of the century. The high marginal price is particularly necessary to reduce emissions from the more nonresponsive sources such as CO 2 emissions from transport or some of the non-co 2 emis-

19 Climatic Change (2007) 81: % 2100 Carbon factor improvement (2000 = 0) 75% 50% 25% Baseline 650 ppm CO 2 -eq. 550 ppm CO 2 -eq. 450 ppm CO 2 -eq. 0% % 25% 50% 75% 100% Energy intensity improvement (2000 = 0) Fig. 6 Relative changes in global energy intensity (energy/gdp) and the carbon factor (CO 2 /energy) in the B2 baseline and the three mitigation cases compared to 2000 values. Note: The diagonal line indicates equal reduction in the energy intensity and carbon factor compared to Values are indicated for all the scenarios: 2020, 2050 and The ovals indicate the outcomes of the mitigation cases for similar years sions from agricultural sources, while other sources, such as electric power, already reduce their emissions to virtually zero at a permit prices of only US$/tC-eq. Costs can also be expressed as abatement costs as a percentage of GDP. This indicator is shown over time (Fig. 7; right panel), and accumulated across the century (net present value; discounted at 5%; Fig. 8). In the 650 ppm CO 2 -eq. stabilization scenario, costs first 2100 Permit price ($/tc-eq) a) Marginal permit price b) Abatement costs Abatement costs (% GDP) ppm CO 2 -eq 550 ppm CO 2 -eq 450 ppm CO 2 -eq Fig. 7 Marginal carbon-equivalent price for stabilizing greenhouse gas concentration at 650, 550 and 450 ppm CO 2 -eq. from the B2 baseline (left; a) and abatement costs as a percentage of GDP for these scenarios (right; b)

20 138 Climatic Change (2007) 81: increase to about 0.5% of GDP, after which they decline slightly to about 0.3% of GDP. This reduction is caused by an increase in global GDP and stabilizing climate costs due to a somewhat lower permit price and a stabilizing emission gap between baseline and the mitigation scenario. The same trend is observed for the other stabilization scenarios, although at higher costs. The abatement costs of the 550 ppm CO 2 -eq. stabilization scenario increase to 1.2% of GDP, while the abatement costs of the 450 ppm CO 2 -eq. stabilization scenario increase to 2.0% of global GDP. The direct abatement costs of about 0% 2.5% of GDP can be compared to the total expenditures of the energy sector (which, worldwide, are about 7.5% of GDP today and expected to remain nearly constant under our baseline) or to the expenditures on environmental policy (in the EU around 2.0% 2.8%, mostly for waste and wastewater management). The net present value of the abatement costs follow a similar trend (across the different stabilization levels) as described above for the costs over time (Fig. 8). For default baseline (B2), the costs vary from 0.2% of GDP for stabilization at 650 ppm to 1.1% of GDP in the 450 case Changes in fuel trade patterns Figure 9 shows the imports and exports of different fuels in The clearest differences are found in the oil and coal trades, which are greatly reduced as a result of lower consumption levels. On the one hand, oil-exporting regions will see their exports reduced by a factor of about 2 3. On the other hand, the oil imports of importing countries are significantly reduced. Interestingly, natural gas trade is hardly affected because natural gas can be used very effectively in combination with CCS. An interesting area is the role played by the bio-energy trade. This trade increases substantially and major exporting regions (including, for instance, South America and the former Soviet Union) could benefit from this. Currently, oil-importing regions (such as the USA, western Europe and Asia) could become major bio-energy importing regions. Fig. 8 Net present value (NPV) of abatement costs for different stabilization levels as percentage of the NPV of GDP, starting from different baseline scenarios (discount rate 5%) NPV Abatement costs (% NPV GDP) 2.0 A1b B2 1.5 B Stabilization level (ppm CO2-eq)

21 Climatic Change (2007) 81: Fig. 9 World volume of fuel trade between the 17 world regions (EJ); 2000, Baseline (B2) and stabilization scenarios (650, 550 and 450 ppm CO 2 -eq) 5 Benefits and co-benefits 5.1 Climate benefits of stabilization The three multi-gas stabilization scenarios analyzed here lead to clearly different temperature increases, both during this century and in the long run. Table 3 shows some of the parameters, describing the different scenarios in more detail and using a single value for climate sensitivity (2.5 C). The table shows that, in 2100, the 650 and 550 ppm CO 2 -eq. stabilization scenarios are still approaching the stabilization levels, while the 450 ppm CO 2 -eq. scenario has in fact overshot its target (as designed) and is approaching its target from a higher concentration level (the 2100 CO 2 -eq. concentration is 479 ppm). For CO 2 only, our three scenarios generate CO 2 concentrations of 524, 463 and 424 ppm for 2100 and this is indeed on the lower side of existing CO 2 -only stabilization scenarios in the literature. It should be noted, however, that the temperature results of the different stabilization scenarios depend to a considerable extent on the uncertain relationship between the GHG concentration and temperature increase. This implies that impacts on temperature can better be expressed in probabilistic terms. Figure 10 shows, on the basis of the work of Meinshausen (2006), the probabilities of overshooting a 2 -C and a 2.5 C target in the light of the Table 3 Overview of several key parameters for the stabilization scenarios explored 2100 concentration (ppm) Reduction of cumulative emissions in period Temperature change C) CO 2 -eq. CO 2 % 2100 Equilibrium B B2 650 ppm CO 2 -eq B2 550 ppm CO 2 -eq B2 450 ppm CO 2 -eq

22 140 Climatic Change (2007) 81: o C temperature target o C temperature target Probability of achieving target CO 2 -eq. concentration (ppm) Probability of achieving target CO 2 -eq concentration (ppm) Fig. 10 Probability of equilibrium temperature change staying within the 2 or 2.5 C limit for compared to pre-industrial for different CO 2 -eq. concentration levels compared to pre-industrial (following calculations of (Meinshausen 2006). Note: The lines indicate the probability function as indicated in the individual studies quoted by (Meinshausen 2006); the grey area indicates the total range between the highest and lowest study different stabilization levels explored in this paper (the corridor shown is a result of the fact that Meinshausen considered several PDFs published in the literature). In the case of a 2 C target, the 650 ppm scenario gives a probability of meeting this target between 0% 18% depending on the PDF used. By contrast, the 450 ppm scenarios result in a probability range of 22% 73%. Similar trends are found for a 2.5 C target. Here, 650 ppm provides a probability range of 0% 37%; 450 ppm a range from 40% 90%. Although we have not specifically targeted any rate of temperature change, a rate can be a useful proxy for the risk of adverse impacts from climate change (in particular ecosystems; see Fig. 11). In the baseline scenario, the rate of temperature change is around 0.25 C per decade. In the mitigation scenarios, the rate of temperature increase drops significantly in particular in the second half of the century. In the 650 ppm stabilization scenario, the rate drops below 0.2 C per decade around 2050 and below 0.1 C in In the 550 and 650 stabilization scenarios, the rate of change drops even further while, for 450 ppm CO 2 -eq., the rate actually falls below zero in In the early decades (until 2030), the mitigation scenarios hardly perform any better than the baseline. The reason is that, in the mitigation scenarios, changes in the energy system to reduce CO 2 emissions also lead to a reduction in sulphur cooling (as already emphasized by Wigley 1991). 15 In our earlier calculations, in fact, this could even lead to an temporarily higher rate of temperature increase for some of our mitigation scenarios compared to baseline (van Vuuren et al. 2006b). The somewhat smaller impact here is mostly due to the increased potential to reduce non-co 2 GHGs, in combination with the higher overall rates of GHG emission reduction. By using GWPs as the basis of substitution between the different greenhouse gases, our method evaluates CH 4 emission reduction as relatively cheap compared to reducing CO 2 (see also van Vuuren et al. 2006c). As reducing CH 4 is much less coupled to reducing sulphur and the impact of reducing CH 4 on radiative forcing is much more direct, the high degree of CH 4 reduction in our scenarios mitigates the impact of reduced sulphur cooling. This is somewhat comparable to the alternative mitigation scenario suggested by Hansen et al. (2000). 15 The impact of sulphur emissions on temperature increase is calculated in IMAGE based on the pattern scaling methodology that was developed by Schlesinger et al. (2000).

23 Climatic Change (2007) 81: Fig. 11 Rate of temperature change for assuming a 2.5 C climate sensitivity Temperature change (Degree C/decade) Baseline 650 CO2-eq 550 CO2-eq 450 CO2-eq Co-benefits and additional costs Impacts on regional air pollutants Many air pollutants and GHGs have common sources. Their emissions interact in the atmosphere and, separately or jointly, cause a variety of environmental effects at the local, regional and global scales. Emission control strategies that simultaneously address air pollutants and GHGs may therefore lead to a more efficient use of resources at all scales. Current studies indicate that, when climate policies are in place, in the short-term (in particular the Kyoto period) potential co-benefits could be substantial, with financial savings in the order of 20% 50% of the abatement costs of the climate policy (e.g., van Vuuren et al. 2006a). In this study, we have focused our analysis on the consequences of climate policies for SO 2 and NO x emissions by using the same emission coefficients for SO 2 and NO x as those assumed under the baseline (reflecting similar policies for emissions of these substances), and simply quantifying the impact of changes in the energy system on emissions. Figure 12 shows that the changes induced by climate policy in the energy system to reduce CO 2 emissions also reduce SO 2 emissions, in particular at lower reduction levels. This can be explained by the fact that coal in particular is used in conventional power plants, Fig. 12 Reduction of CO 2 emission compared to baseline (baseline = 100%) in the three B2 stabilization scenarios vs reductions of SO 2 and NO x emissions compared to baseline (2050 on left; 2100 on right)